Multi-objective decision in machine learning.

Abstract

Thiswork presents a novel approach for decisionmaking for multi-objective binary classification problems. The purpose of the decision process is to select within a set of Pareto-optimal solutions, one model that minimizes the structural risk (generalization error). This new approach utilizes a kind of prior knowledge that, if available, allows the selection of a model that better represents the problem in question. Prior knowledge about the imprecisions of the collected data enables the identification of the region of equivalent solutions within the set of Pareto-optimal solutions. Results for binary classification problems with sets of synthetic and real data indicate equal or better performance in terms of decision efficiency compared to similar approaches.